Start fresh - merge the loom files

Man I hate to do it over and over again.

Hope this time the data looks better!

Interesting links to B and T cell development: https://www.sciencedirect.com/science/article/pii/S1074761319304042?via%3Dihub https://en.wikipedia.org/wiki/V(D)J_recombination

These clusters are not OK

The clusters in 3D starting with louvain lcuistersin resolution 11 and merging down with 0.95 look way better. Instead of using the 'old' clustering I will try whether using the same settings does reproduce the clsutering here. It really should!

YES!!!

This method is totally reproducable and it does not make a differnece if one starts with 2D or 3D data. Apart from the data looking much better in 3D :-D

Short explanation of the mature B cell idea:

I am surprised that the B cells cluster by both heavy chain usage (IGHA and IGHG) as well as light chain usage (IGLC and IGKC). If there is so much difference in these cells what are the differences?

Short the eraly B/T cell development

I will refere to group IDs as in 'louvain11_merged.95_renamed'. group 19 is the central group where both early B and T cells have a connection to. Hence I assume it is the CLP population. Groups 2 and 32 seam to be early B cells (proB; heavy chain recombination) and group 1 are preB cells (light chain VDJ recombination). They all express low level of IGHM. Even parts of group 7 express IGHM and link to the also IGHM expressing groups 0, 21 and 24. I do not know which cells these are. The group 34 even expressed IGHD - a marker for immature B cells.

The interesting parts are (A) why are there T cells in the bone marrow (groups 11 and 17) and (B) is there really a connection between the proB group 32 and the T cell group 11? That would be cool to see as some T cells are also harboring IgHM recombinations - failed ones. But proB should be already committed to B cell development. So how do these cells manage to drop out of the B cell program?

A shame I do not have enough time to dig deeper.

Do it yourself.

Hi Hongzhe,

I have added a folder on Aurora for you that contains all the overclustered (finaly) analysed file.

/projects/fs1/common/Hongzhe/Hongze_healthy_2020_10_downsampled_1000_2D_overclustered.h5ad

Copy this file to a folder you want to work in (on aurora), go to this folder (terminal) and run

sbatch /projects/fs1/common/software/SingSingCell/1.0/runSbatch.sh

You get a jupyter notebook server that runs for 24h on one node. You will get a message like

Submitted batch job 746197

by running

cat <the id you saw here 746197>

You get to know how to connect to the server:

[InstallKernelSpec] Removing existing kernelspec in /home/stefanl/.local/share/jupyter/kernels/ir
[InstallKernelSpec] Installed kernelspec ir in /home/stefanl/.local/share/jupyter/kernels/ir
[I 08:25:10.625 LabApp] JupyterLab extension loaded from /usr/local/lib/python3.8/dist-packages/jupyterlab
[I 08:25:10.625 LabApp] JupyterLab application directory is /usr/local/share/jupyter/lab
[I 08:25:10.627 LabApp] Serving notebooks from local directory: /projects/fs1/common/genome/lunarc/datasets/A_human_cell_atlas_of_fetal_gene_expression
[I 08:25:10.628 LabApp] Jupyter Notebook 6.1.4 is running at:
[I 08:25:10.628 LabApp] http://ls2-n3:9734/?token=1a83ddcf514656277436b6dd191d56b948cbb7a828e02cac
[I 08:25:10.628 LabApp]  or http://127.0.0.1:9734/?token=1a83ddcf514656277436b6dd191d56b948cbb7a828e02cac
[I 08:25:10.628 LabApp] Use Control-C to stop this server and shut down all kernels (twice to skip confirmation).

There you riught click on the line like -> open link

[I 08:25:10.628 LabApp] http://ls2-n3:9734/?token=1a83ddcf514656277436b6dd191d56b948cbb7a828e02cac

In this jupyter labbook server you can plot after:

Hope you enjoy this experience. Feel free to try whatever you want!